The structure of scientiﬁc collaborations has been the object of intense study both forits importance for innovation and scientiﬁc advancement, and as a model system forsocial group coordination and formation thanks to the availability of authorship data. Over the last years, complex networks approach to this problem have yielded important insights and shaped our understanding of scientiﬁc communities. In this paper we propose to complement the picture provided by network tools with that coming from using simplicial descriptions of publications and the corresponding topological methods. We show that it is natural to extend the concept of triadic closure to simplicial complexes and show the presence of strong simplicial closure. Focusing on the diﬀerences between scientiﬁc ﬁelds, we ﬁnd that, while categories are characterized by diﬀerent collaboration size distributions, the distributions of how many collaborations to which an author is able to participate is conserved across ﬁelds pointing to underlying attentional and temporal constraints. We then show that homological cycles, that can intuitively be thought as hole in the network fabric, are an important part of the underlying community linking structure.